MineBench
MineBench is an AI voxel build benchmark for comparing model performance on text prompts. It uses curated natural-language prompts to probe spatial reasoning, then asks models to output raw JSON coordinates for voxel blocks. The site renders those block coordinates directly, with no post-processing, and presents the resulting builds for comparison.
The benchmark workflow is built around head-to-head evaluation. Two builds are shown side by side, humans vote pair-wise, and rankings emerge from Elo. The interface includes options such as A wins, B wins, Tie, Bad, and Skip, and the leaderboard updates as votes are added. MineBench also includes a Sandbox for entering any prompt and generating a 3D build for testing.
MineBench is delivered as a web-based product at minebench.ai. The site also links to Arena, Sandbox, Local Leaderboard, Support, and a Buy Me a Coffee support option. Its page describes the project as an AI spatial reasoning benchmark and as a Minecraft-style voxel build benchmark, which reflects the format of the builds and the comparison-based evaluation approach.
FreeMITWebCloud-managed
271stars
18forks
5features
2026since
Overview
5 featuresMineBench is a LLM eval & observability project. It focuses on comparing and evaluating AI models' spatial reasoning and generation capabilities in a standardized benchmark. MineBench is a B2B product aimed at ai researchers. MineBench costs nothing to use. It runs on the web.
MineBench first shipped in 2026. Development happens publicly on GitHub with 271 stars and 210 commits in the last 90 days. Among its 5 catalogued features are voxel build benchmark, AI model comparison, and live leaderboard.
Summary written by a language model from the project’s public pages.
- Voxel build benchmark
- AI model comparison
- Live leaderboard
- Prompt-based evaluation
- User voting
Tags
ai-benchmarkvoxel-evaluationmodel-leaderboard
AI capabilities
Structured
Built with & integrations
Framework
Next.jsnext
Hosting
Vercel
AI providers
multipleopenai
Runs on
BrowserCloud-managed
Detected from
- Next.js
- x-nextjs-prerender header · /_next/static/ in the HTML · __next_f in the HTML
- openai
- bgpt- in the HTML
- Vercel
- x-vercel-id header · x-vercel-cache header
Trust & compliance
License
MIT
Verified signals
Indexing history
1What PulseGate has recorded for this listing
Frequently asked questions about MineBench
- What is MineBench?
- MineBench focuses on comparing and evaluating AI models' spatial reasoning and generation capabilities in a standardized benchmark. It is catalogued under LLM eval & observability on PulseGate.
- Who is MineBench for?
- MineBench is a B2B product built for ai researchers.
- Is MineBench free?
- Yes — MineBench is free to use.
- What platforms does MineBench run on?
- MineBench runs on the web.
- Is MineBench still active?
- The GitHub repository shows 210 commits in the last 90 days.
- What are alternatives to MineBench?
- Similar projects tracked by PulseGate include Bench'd, MazeBench, and YourBench.Bench'dMazeBenchYourBench
- When did MineBench launch?
- MineBench first shipped in 2026.
- Is MineBench open source?
- MineBench has a public GitHub repository.
Similar projects
Closest matches by what these projects do